Organisations are not usually short of data. In high-hazard industries, there are often incident reports, near-miss records, audits, observations, investigations, KPIs, procedure reviews and safety dashboards.
These sources of information are useful for a number of reasons. They help organisations understand performance, identify where things have gone wrong, and spot where performance may be drifting. They can also support regulatory and company requirements by providing a record of historical events and decisions.
However, data only becomes valuable when the organisation learns something from it.
Does the data help explain why work happened in the way it did? Does it reveal the conditions that made error more likely? Does it challenge assumptions about how tasks are performed? Does it lead to changes in procedures, training, job aids, supervision or work design?
This is where Human Factors becomes essential. Rather than treating human error as the end point for any incidents or near misses, a Human Factors approach asks what shaped people’s performance. It looks at the wider system: procedures, training, equipment, time pressure, communication, staffing, organisational priorities and the culture in which work takes place.
In other words, it helps organisations move from collecting information about failure to learning from work as it is actually done.
Data Collection Is Only The Starting Point
When we think of data, we often think about numbers, percentages and trends. However, data can also include findings from observations, interviews, investigations, and operational feedback.
The purpose of data collection should not simply be to document what happened. It should be to understand why it happened, and what needs to change to reduce the likelihood of recurrence.
Although data collection often takes place after an accident or near miss, it can still play a proactive role in accident prevention. When organisations look beyond the immediate event and examine underlying causes, they can identify conditions and vulnerable areas that may lead to future incidents before those incidents occur.
A weak data collection system may stop at the visible event: someone missed a step, selected the wrong item, responded too slowly, or did not follow the procedure. These descriptions may be factually true, but they rarely explain the conditions that made the action more likely at the time.
A stronger system asks deeper questions. Was the procedure accurate and usable? Was the task being carried out under time pressure? Was the person trained for the situation they faced? Was the equipment designed in a way that supported correct action? Was there a conflict between production and safety? Were local workarounds already normal practice?
In this sense, data collection should be seen as the beginning of a learning process, not the end of an administrative one.
The Organisation’s View Of Human Error Shapes What It Learns
Organisations often believe they are collecting objective data, but what they choose to collect could be shaped by what they believe is important.
If the starting assumption is that people cause accidents because they are careless, negligent or deliberately non-compliant, then investigations are likely to focus on individuals and unsafe acts. The organisation may collect data that reinforces this view, leading to actions such as reminders, increased supervision, disciplinary responses or motivational campaigns.
These responses may sometimes have a place, but they can also miss the wider issues. Procedures may not match the realities of the task. Training may not prepare people for the conditions they actually face. Equipment may make the correct action difficult. Time pressure may encourage shortcuts. Organisational priorities may unintentionally reward risky ways of working.
A systems approach changes the focus. It does not ignore individual actions, but places them in context. Human performance is shaped by the design of the work system, and data collection should help reveal that system. This includes the task, tools, environment, procedures, culture and organisational decisions that create the conditions for success or failure.
This is where the difference between collecting and learning becomes visible. Data collection records what the organisation has chosen to look for. Learning begins when the organisation is willing to question whether it has been looking in the right place.
Culture Determines Whether Data Is Useful
While collecting the correct type of data is helpful for learning, even the best-designed data collection system will fail if people do not trust it.
If people believe that reporting an error, near miss or workaround will lead to blame, punishment or reputational damage, they are less likely to report openly. This is especially important when organisations are trying to understand weak signals, informal adaptations, or the early warning signs of risk.
A reporting system can look impressive on paper and still provide a limited view of reality. If workers only report what feels safe to report, the organisation may receive a sanitised version of work. It may see the events that are formally captured, but not the everyday difficulties, trade-offs and adaptations that shape performance.
For data collection to support learning, people need confidence that the information will be used constructively. This means having a non-punitive culture, freedom from fear of reprisals, and visible evidence that feedback leads to positive change. Without this, people can become disengaged, and the system becomes less useful.
A learning culture therefore depends on more than a reporting form. It depends on trust, visible action, and a shared belief that understanding real work is more valuable than assigning blame.
The missing link: Work-As-Imagined and Work-As-Done
In high-risk industries, procedures are often treated as a key part of the safety management system. They are expected to define safe methods, support quality, provide consistency and demonstrate control. However, there is often a gap between formal written procedures and the way work is actually carried out.
This gap matters because procedures are often assumed to represent how risk is being controlled. If the procedure describes one version of work, but people routinely perform the task in another way, then the organisation may not have an accurate picture of how risk is really being managed.
Importantly, this gap is not automatically a sign of poor behaviour. In many cases, workers adapt because the formal procedure does not fully reflect the realities of the task. Their adaptations may be sensible, efficient and safe. They may be the reason the system continues to function despite imperfect conditions.
Procedures may drift away from practice for several reasons. They may be written by people who understand the technical system, but have less direct experience of the practical constraints of the work environment. They may not be updated when plant conditions, equipment, staffing or working practices change. They may be too complex, too restrictive, difficult to find or unrealistic under operational pressures.
When this happens, procedures lose credibility. Instead, people may rely on memory, experience, informal notes, local workarounds, or knowledge passed between colleagues. In some cases, this informal knowledge may be highly valuable. In others, it may create inconsistency and hidden risk.
The issue is not simply “procedure compliance”. It is whether the procedure genuinely reflects the safest and most practical way of doing the work.
This makes procedure management a learning issue, not just a documentation issue. To manage risk effectively, organisations need to understand both Work-As-Imagined and Work-As-Done, and then use that understanding to keep procedures, training, job aids and competency systems aligned with real work.
Turning Operational Experience into Best Practices
The Consensus-based Approach to Risk Management (CARMAN) provides one practical way to close the gap between formal procedures and actual working practices.
CARMAN involves the people who carry out the work in understanding current practice, evaluating risk and developing agreed best practices. This matters because the people closest to the task often hold knowledge that is not visible in formal systems, including where procedures are difficult to apply, where work varies between shifts, and where informal workarounds have developed.
Through facilitated consensus groups, operators, technicians, technical specialists and managers can compare current working practices, discuss hazards and consequences, and agree the safest and most practical way to carry out the task. The aim is not simply to rewrite procedures for the workforce, but to develop best practices with the workforce.
This will help bring best practices and preferred practices closer together. A procedure may describe the “right” way to do the task, but if it is not seen as practical, credible or useful, it is unlikely to become the normal way of working. Equally, a preferred way of working may be practical and efficient, but if it has not been assessed for risk, it may not be safe or consistent enough to become best practice.
By involving workers in the process, organisations can create shared ownership of an agreed approach that people understand, trust and are more likely to use. Procedures then become a practical expression of collective knowledge about how to do the job safely and effectively.
This also helps align the wider systems that support performance. In many organisations, procedures, training, job aids and competency assessment can become disconnected. A more effective approach is to use agreed Best Practice as the foundation for these elements, creating a more consistent understanding of what good performance looks like and what support people need.
It also allows organisations to be more intelligent about procedural support. Not every task needs a detailed step-by-step instruction. Some tasks may be performed primarily using skill and experience, particularly when they are familiar and lower risk. Other tasks, especially those that are critical, complex or infrequently performed, may require job aids, checklists, flowcharts, memory aids or detailed instructions.
Good support is not about producing more documentation. It is about providing the right information, in the right format, at the right time.
Closing the Human Factors learning loop
Taken together, data collection and consensus-based risk management describe a wider learning loop.
Data collection provides information about events, near misses and underlying causes, and a consensus-based approach provides a way to turn operational knowledge into shared best practices.
The loop looks something like this:
- Collect data from incidents, near misses and operational experience.
- Understand the underlying causes and system conditions.
- Identify the gap between Work-As-Imagined and Work-As-Done.
- Involve the people who understand the task.
- Agree practical, risk-informed best practices.
- Update procedures, training, job aids and competency standards.
- Gather feedback from operational experience.
This is organisational learning in practice. It is not a one-off investigation, a report, or a database entry. It is an ongoing process of using information from real work to improve how work is designed, supported and managed.
Importantly, organisations should not only learn after something has gone wrong. Normal work also contains rich information about how people keep systems functioning. It shows how they adapt to changing conditions, recover from small disturbances, manage competing demands and compensate for weaknesses in the system.
These everyday adjustments may not appear in incident statistics, but they are central to how safety and performance are actually achieved. By understanding normal work, organisations can identify where resilience exists, but also where the system may be relying too heavily on individual expertise or informal workarounds. This creates an opportunity to strengthen the system before failure occurs.
Conclusion
Collecting data is not the same as learning from work.
Data becomes valuable when it helps organisations understand and improve the conditions that shape human performance. Having a consensus approach, involving people who do the job, provides one way to turn operational experience into practical, risk-informed best practice.
Together, they show that Human Factors is not just about analysing error after the event. It is about closing the loop between Work-As-Imagined, Work-As-Done and having continuous improvement.
This blog revisits two HRA resources from our archive: Data Collection Systems and Preventing Human Error: Developing a Consensus Led Safety Culture based on Best Practice.
The language and context may reflect the time in which they were written, but the core ideas remain highly relevant till this day.
